Masuo Takigawa

Transient exhaust gas improvement by adaptive neural network

Abstract Three-way catalyzer (TWC) is the most common exhaust gas treatment device for gasoline engines. A/F must be, however, kept within very narrow range. The conventional engine control system can maintain this by OZ feedback in a steady state, but not in a transient state. To overcome this, feed-forward control was provided using the neural network (NN) which is suitable to nonlinear behavior. Moreover, the NN has an adaptive ability by providing an on-line backpropagation loop, so that it can deal with different characteristics of same-type engines, aging in the same engine, and so on. In the experiments, relatively good results were obtained.

Modeling and control of electrically assisted power steering.

This paper outlines the development of a simulation method which predicts the handle torque of an electrically assisted power steering system. The electrically assisted power steering system would provide benefits for fuel economy and flexibility in controlling the handle torque. A DC motor with torque sensory feedback is used in this system. However, the DC motor tends to cause torque fluctuation and the feel of rough handling. In order to reduce torque fluctuation, an equation of motion with a 5-degree of freedom model has been formulated for analysis on the electrically assisted power steering, and the methods of controlling the handle torque have been investigated. Feasibility experiments and simulation results show that the simulation method is effective for developing an electrically assisted power steering system.